A multi-modal physiological signal coupling analysis method, system, terminal and storage medium

Abstract

The application relates to the technical field of biological signal processing, and discloses a multi-modal physiological signal coupling analysis method, a system, a terminal and a storage medium. The method comprises the following steps: synchronously collecting multi-modal physiological signals of a subject when the subject performs a specific experimental paradigm; constructing a multi-modal dynamic causal model comprising a neuron dynamics model and an observation model corresponding to each signal; adopting a staged Bayesian inversion strategy to perform parameter estimation, fixing a first type of signal parameter to invert a second type of signal related parameter, fixing the inverted parameter to invert the first type of signal parameter, and obtaining a joint posterior distribution; and finally generating a biomarker of the subject based on the model parameters obtained through inversion. Through the staged inversion strategy, the technical problems of high computational complexity and unstable parameter estimation caused by the large time scale difference of multi-modal data and the large number of model parameters are effectively solved, and the biomarker can be generated and used for clinical motor function evaluation.

Description

Technical Field

[0001] This invention relates to the field of biological signal processing technology, and in particular to a multimodal physiological signal coupling analysis method, system, terminal, and storage medium. Background Technology

[0002] With the development of physiological signal acquisition technology, it has become possible to simultaneously acquire multiple modalities of physiological signals, such as electroencephalogram (EEG), functional near-infrared spectroscopy (fNIRS), and surface electromyography (sEMG). This provides a valuable data foundation for a deeper understanding of the interactions between systems such as the nervous system, blood vessels, and muscles. Dynamic causal modeling (DCM), as a powerful Bayesian framework, can infer effective connections (i.e., causal effects) within or between systems from time-series data, making it a powerful tool for analyzing the coupling relationships of multimodal physiological signals.

[0003] However, applying dynamic causal models to multimodal physiological signal analysis faces significant challenges. First, the vast differences in timescales between different modalities (e.g., millisecond-level EEG or surface electromyography signals versus second-level functional near-infrared spectroscopy signals) lead to complex dynamic processes and high dimensionality in the parameter space of a unified model. Second, the large number of model parameters makes traditional single-stage inversion methods computationally extremely complex and prone to getting trapped in local optima, resulting in poor accuracy and robustness in parameter estimation. Furthermore, most existing technologies only provide statistical correlation indicators, lacking quantitative and interpretable biomarkers with clear physiological significance that can be directly used for clinical assessment.

[0004] Therefore, existing technologies still need to be improved and developed. Summary of the Invention

[0005] The main objective of this invention is to provide a multimodal physiological signal coupling analysis method, system, terminal, and computer-readable storage medium to solve the technical problems of high computational complexity and unrobust parameter estimation caused by large differences in the time scales of multimodal data and numerous model parameters in the prior art. To achieve the above objective, this invention provides a multimodal physiological signal coupling analysis method, which includes the following steps:

[0006] Simultaneously collect physiological signals from subjects in multiple modalities while performing a specific experimental paradigm;

[0007] A multimodal dynamic causal model is constructed, which includes a neuronal dynamics model and observation models corresponding to various physiological signals respectively;

[0008] Based on the physiological signals, a staged Bayesian inversion strategy is used to invert and estimate the parameters of the multimodal dynamic causal model to obtain the model parameters;

[0009] Based on the model parameters obtained from the inversion, biomarkers for the subject are generated;

[0010] The various modal physiological signals include first-type physiological signals and second-type physiological signals.

[0011] Furthermore, based on the physiological signals, a staged Bayesian inversion strategy is used to invert and estimate the parameters of the multimodal dynamic causal model to obtain the model parameters, including:

[0012] By fixing the observation model parameters corresponding to the first type of physiological signal, and using the observation data of the second type of physiological signal, some or all of the parameters of the neuronal dynamics model and the observation model parameters of the second type of physiological signal are inverted and estimated.

[0013] By fixing the parameters obtained from the first-stage inversion estimation, the observation model parameters of the first type of physiological signal are inverted and estimated using the observation data of the first type of physiological signal to obtain the joint posterior distribution of all parameters.

[0014] Furthermore, the first type of physiological signal is a low-frequency signal with a time resolution lower than a preset threshold, and the second type of physiological signal is a high-frequency signal with a time resolution higher than the preset threshold;

[0015] The low-frequency signal includes functional near-infrared spectral signals, and the high-frequency signal includes electroencephalogram (EEG) signals and surface electromyography (EMG) signals.

[0016] Furthermore, the neuronal dynamics model is a standardized microcircuit neural cell group model, used to simulate neural electrical activity within and between brain regions.

[0017] Furthermore, the standardized microcircuit neural cell population model includes four types of neuronal cell populations distributed in the granular layer, upper granular layer, and lower granular layer. These four neuronal cell populations include excitatory polyspinous stellate cell populations, superficial pyramidal cell populations, deep pyramidal cell populations, and inhibitory interneuronal cell populations. The intra-brain region connectivity between these neuronal cell populations is described by endogenous connectivity parameters, while the inter-brain region connectivity between models of different brain regions is described by exogenous connectivity parameters.

[0018] Furthermore, the observation models corresponding to each of the various physiological signals include:

[0019] An EEG signal observation model is used to convert transmembrane potentials generated by neuronal activity into scalp potential signals through a linear mapping matrix.

[0020] A functional near-infrared spectral signal observation model is used to describe the changes in cerebral blood flow and hemoglobin concentration caused by vasodilation induced by neuronal activity.

[0021] A surface electromyography (EMG) signal observation model is used to convert descending nerve drive signals from the motor cortex into EMG signals recorded on the skin surface.

[0022] The parameters to be estimated in the surface electromyography signal observation model include basis function weight parameters and neuromuscular conduction delay parameters.

[0023] Furthermore, the phased Bayesian inversion strategy employs the variational Laplace algorithm, which maximizes the negative variational free energy of the model by iteratively updating the mean and covariance of the posterior distribution of the parameters.

[0024] Furthermore, to achieve the above objectives, the present invention also provides a multimodal physiological signal coupling analysis system, wherein the multimodal physiological signal coupling analysis system comprises:

[0025] The signal acquisition module is used to simultaneously acquire physiological signals from multiple modalities of the subject while performing a specific experimental paradigm;

[0026] The model building module is used to build a multimodal dynamic causal model, which includes a neuronal dynamics model and observation models corresponding to various physiological signals.

[0027] The inversion estimation module is used to invert and estimate the parameters of the multimodal dynamic causal model based on the physiological signals using a staged Bayesian inversion strategy, so as to obtain the model parameters.

[0028] A biomarker generation module is used to generate biomarkers for the subject based on the model parameters obtained by inversion.

[0029] In addition, to achieve the above objectives, the present invention also provides a terminal, wherein the terminal includes: a memory, a processor, and a multimodal physiological signal coupling analysis program stored in the memory and executable on the processor, wherein when the multimodal physiological signal coupling analysis program is executed by the processor, it implements the steps of the multimodal physiological signal coupling analysis method as described above.

[0030] In addition, to achieve the above objectives, the present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multimodal physiological signal coupling analysis program, which, when executed by a processor, implements the steps of the multimodal physiological signal coupling analysis method as described above.

[0031] The beneficial effects of this invention are as follows: By constructing a unified model framework and adopting a phased inversion strategy, this invention decomposes the originally highly complex computational problem into manageable and sequentially solvable sub-problems. This enables in-depth analysis of complex physiological systems to be completed within reasonable computational resources and time. The "phased Bayesian inversion strategy" of this invention utilizes the advantages of different modal physiological signals, complementing and constraining each other, greatly reducing the uncertainty of parameter estimation and improving the reliability and repeatability of the results. Therefore, this invention effectively solves the technical challenges of high computational complexity and unrobust parameter estimation caused by large differences in the time scale of multimodal data and numerous model parameters. Attached Figure Description

[0032] Figure 1 This is a flowchart of a preferred embodiment of the multimodal physiological signal coupling analysis method of the present invention;

[0033] Figure 2 This is a schematic diagram of the standard microcircuit neural cell population model described in this invention;

[0034] Figure 3 This is a structural diagram of a preferred embodiment of the multimodal physiological signal coupling analysis system of the present invention;

[0035] Figure 4 This is a structural diagram of a preferred embodiment of the terminal of the present invention. Detailed Implementation

[0036] This application provides a method, system, terminal, and storage medium for multimodal physiological signal coupling analysis. To make the objectives, technical solutions, and effects of this application clearer and more explicit, the following detailed description is provided with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit this application.

[0037] It will be understood by those skilled in the art that, unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0038] Furthermore, if the embodiments of this invention involve descriptions such as "first" or "second," these descriptions are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first" or "second" may explicitly or implicitly include at least one of those features. Additionally, the technical solutions of the various embodiments can be combined with each other, but this must be based on the ability of those skilled in the art to implement them. If the combination of technical solutions is contradictory or impossible to implement, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by this invention.

[0039] The multimodal physiological signal coupling analysis method described in the preferred embodiment of the present invention, such as... Figure 1 As shown, the multimodal physiological signal coupling analysis method includes the following steps:

[0040] S10. Simultaneously collect physiological signals from subjects in multiple modalities while performing a specific experimental paradigm.

[0041] Furthermore, the multiple modal physiological signals include signals with different time resolutions. The first type of physiological signal is a low-frequency signal with a time resolution lower than a preset threshold, and the second type of physiological signal is a high-frequency signal with a time resolution higher than the preset threshold.

[0042] Furthermore, the low-frequency signal includes functional near-infrared spectral signals, and the high-frequency signal includes electroencephalogram (EEG) signals and surface electromyography (EMG) signals.

[0043] Specifically, based on the specific pattern of motor dysfunction in the subject (e.g., a patient), a corresponding motor experiment paradigm is designed. For example, tasks such as isokinetic contraction of the upper limbs or knee flexion and extension of the lower limbs can be designed. According to the experimental paradigm, the brain region of interest (ROI) is identified, and EEG electrodes, as well as light sources and detectors for functional near-infrared spectroscopy (FIR) signals, are deployed. Simultaneously, the muscle groups to be monitored are identified, and surface electromyography (EMG) electrodes are attached. Using a multimodal physiological signal synchronous acquisition device, the subject is guided to perform the predetermined paradigm movements, and EEG signals, FIR signals, and surface EMG signals are recorded simultaneously during the movement.

[0044] S20. Construct a multimodal dynamic causal model, which includes a neuronal dynamics model and observation models corresponding to various physiological signals respectively; further, the neuronal dynamics model is a standardized microcircuit neural cell group model, used to simulate neural electrical activity within and between brain regions.

[0045] Step S20 specifically includes the following steps:

[0046] Step S21: Determine the neuronal dynamic model: such as Figure 2As shown, a canonical microcircuit neural mass model is used as a unified neuronal dynamics model to simulate neural electrical activity within and between brain regions. This model consists of four neuronal cell populations distributed in the granular, upper, and lower granular layers: excitatory polyspinous stellate cells, superficial pyramidal cells, deep pyramidal cells, and inhibitory interneurons. These cell populations are interconnected via endogenous connectivity parameters (…). a ij ) describes the connectivity within brain regions, and models of different brain regions are connected through exogenous connectivity parameters ( A ij The diagram describes the connectivity between brain regions. Here, i represents the target cell population (i = 1, 2, 3, or 4), and j represents the source cell population (j = 1, 2, 3, or 4). Experimental stimuli directly drive neural activity through weighted coefficients (C) or indirectly modulate connectivity parameters through a modulation matrix (B). Neuronal activity is described by transmembrane potentials (V), and their dynamics are characterized by second-order ordinary differential equations, as follows:

[0047] ;

[0048] in, The average transmembrane potential of the cell population is expressed as a result of experimental stimulation. The second derivative of the average transmembrane potential of the cell population. The first derivative of the average transmembrane potential of the cell population. T The synaptic rate constant represents the decay rate of the postsynaptic current and controls the response speed of the neuron to the input signal. , and These represent the endogenous inhibitory afferent signal, the endogenous excitatory afferent signal, and the exogenous afferent signal, respectively. These three afferent signals are calculated from the corresponding connection parameters and the Sigmoid function.

[0049] The Sigmoid function is: ;

[0050] in, is the activation function for membrane potential, used to convert an arbitrary input value (e.g., the neuron's "membrane potential") into a bounded, smooth output signal (e.g., the neuron's "firing rate" or "pulse firing probability"). It is a natural constant, approximately equal to 2.71828.

[0051] It should be noted that, in Figure 2 In, such as a 42 ,a 21 , a 12 These are all the intrinsic connection parameters ( a ij ) specific display, such as A 24 , A 34 , A 12 All of these are the exogenous connection parameters ( A ij The specific data format of ); C 11 G1 represents the weighted coefficient of the driving force of the experimental stimulus on the activity of the 11th cell population; G2, G3 and G4 represent the inhibitory connectivity strength of the polyspinous stellate cell population, the superficial pyramidal cell population, the interneuronal cell population and the deep pyramidal cell population, respectively.

[0052] Step S22: Establish a multimodal physiological signal observation model.

[0053] The multimodal physiological signal observation model includes:

[0054] An electroencephalogram (EEG) signal observation model is used to convert transmembrane potentials generated by neuronal activity into scalp potential signals through a linear mapping matrix. The parameters to be estimated in the EEG signal observation model are the position and orientation of the equivalent current dipole. The expression of the EEG signal observation model is:

[0055] ;

[0056] in, Indicates at a point in time Electroencephalogram (EEG) signals measured on scalp electrodes; Indicates at a point in time The average membrane potential of a population of neurons in the cerebral cortex; The weighted matrix representing the contribution of each cell population to the overall transmembrane potential output is a pre-set empirical value; Indicates observation noise; The guiding field is defined by the parameters of the EEG signal observation model. It is an equivalent current dipole.

[0057] The expression for the equivalent current dipole is:

[0058] ;

[0059] in, and These represent the position and orientation attributes of the equivalent current dipole, respectively.

[0060] A functional near-infrared spectroscopy signal observation model is used to describe the changes in cerebral blood flow and hemoglobin concentration caused by vasodilation induced by neuronal activity.

[0061] The observation equation of the functional near-infrared spectral signal observation model can be expressed as:

[0062] ;

[0063] in, It is the collected functional near-infrared spectral signal. This represents the change in relative oxygenated hemoglobin concentration. This represents the relative change in deoxyhemoglobin concentration; the relative amount is relative to the sampling time of 0.

[0064] The expressions for the relative change in oxygenated hemoglobin concentration and the relative change in deoxygenated hemoglobin concentration are as follows:

[0065] and ;

[0066] in, Resting blood volume is usually a fixed parameter, generally around 0.04; , and These represent the constants related to the optical path length and absorption coefficient, respectively. They are related to the wavelength used in the functional near-infrared spectral signal and remain fixed during the calculation. They can be calculated using a modified Lambert-Beer law. Indicates the normalized venous deoxyhemoglobin content; This is the normalized venous volume.

[0067] The nonlinear equation for the venous deoxyhemoglobin content is:

[0068] ;

[0069] in, It is the normalized venous deoxyhemoglobin content. It is the first derivative of the normalized venous deoxyhemoglobin content; This is the hemodynamic transport rate constant, representing the time delay of blood flow regulation of blood volume; For normalized cerebral blood flow, The resting oxygen uptake fraction is generally fixed at an empirical value of around 0.4. Normalized venous volume, The Grubb exponent is a parameter describing the nonlinear power-law relationship between cerebral blood flow and cerebral blood volume, typically 0.32. This equation simulates the balance between oxygen brought in by blood flow and oxygen consumed by neural activity: increased blood flow dilutes deoxyhemoglobin, leading to a decrease in deoxyhemoglobin levels, thereby generating a functional near-infrared spectral signal.

[0070] The differential expression for the venous vessel volume is:

[0071] ;

[0072] in, The first derivative of the normalized venous vessel volume, This is the normalized venous volume.

[0073] The rate of change of cerebral blood flow is determined by vasodilation signals, and the expression for the rate of change of cerebral blood flow is:

[0074] ;

[0075] in, The differential of cerebral blood flow This indicates the degree of relaxation of arterioles and precapillary sphincters. It should be noted that cerebral blood flow needs to be normalized, and is 1 in the resting state.

[0076] The expression for the degree of relaxation of the arterioles and precapillary sphincters is as follows:

[0077] ;

[0078] in, express The differential, This is a parameter representing the neurovascular coupling efficiency, indicating the intensity of vasodilation driven by neural activity; The driving neural activity input, in this invention, is the total output of the neuronal dynamics model, that is, the sum of all presynaptic inputs; The signal attenuation rate constant represents the attenuation rate of the vasodilatory signal itself. The feedback rate constant is a blood flow-dependent constant, representing the intensity of negative feedback regulation of cerebral blood flow (CBF) on vasodilation signals (based on the blood flow autoregulation mechanism). Normalized cerebral blood flow.

[0079] It should be noted that the parameters to be estimated in the functional near-infrared spectral signal observation model include... , , , and .

[0080] A surface electromyography (EMG) signal observation model is used to convert descending nerve drive signals from the motor cortex into EMG signals recorded on the skin surface. The expression for the surface EMG signal observation model is:

[0081] ;

[0082] in, These are surface electromyographic signals recorded on the skin surface. The linear transmission nucleus represents the total effect of the entire pathway from "the brain issuing a command" to "the muscle generating an electrical signal"; The descending driving signal of the cerebral motor cortex is the input of the surface electromyography signal observation model, which originates from the excitation rate of deep pyramidal cells in the primary motor cortex in the standardized microcircuit neural cell group model. These cells are the main origin of the corticospinal tract. Indicates the sampling time. Neuromuscular conduction delay represents the time from when the brain issues a command to when the muscle generates a detectable electrical signal, typically between 20 and 50 ms.

[0083] The expression for the linear transfer kernel is:

[0084] ;

[0085] in, For linear transfer kernel, This represents the total number of basis functions (components). For the first Several basis functions are used to reflect physiological characteristics, and the gamma function is used to simulate the surface electromyographic signal impulse response generated by the activation of a motor unit: The weight parameters for each basis function represent the strength or gain of the pathway; The parameters of the basis functions, such as shape or scale parameters, are generally preset.

[0086] It should be noted that the parameters to be estimated in the surface electromyography signal model include: basis function weight parameters and neuromuscular conduction delay parameters.

[0087] Step S23: Integrate and construct a multimodal dynamic causal model (DCM): Based on a defined region of interest (ROI), including the left / right prefrontal cortex and the left / right sensorimotor cortex; instantiate a neuronal dynamics model for each region, and construct a neuronal network through exogenous feedforward connections (e.g., from the prefrontal cortex to the sensorimotor cortex) and exogenous feedback connections (e.g., from the sensorimotor cortex to the prefrontal cortex). This neuronal network, together with the aforementioned EEG signal, functional near-infrared spectroscopy signal, and surface electromyography signal observation models, constitutes the multimodal dynamic causal model; then, set a reasonable prior probability distribution (e.g., Gaussian distribution) for all model parameters (including connection parameters and observation model parameters).

[0088] S30. Based on the physiological signals, a staged Bayesian inversion strategy is used to invert and estimate the parameters of the multimodal dynamic causal model to obtain the model parameters. Specifically, this includes:

[0089] First-stage inversion: Fix the observation model parameters corresponding to the first type of physiological signal, and use the observation data of the second type of physiological signal to invert and estimate some or all of the parameters of the neuronal dynamics model and the observation model parameters of the second type of physiological signal;

[0090] Second-stage inversion: Fix the parameters obtained by the first-stage inversion estimation, and use the observation data of the first type of physiological signal to perform inversion estimation on the parameters of the first type of physiological signal observation model to obtain the joint posterior distribution of all parameters;

[0091] Furthermore, the first stage of inversion specifically involves: fixing the parameters of the functional near-infrared spectral signal observation model and performing inversion using observation data of electroencephalogram (EEG) signals and surface electromyography (EMG) signals; the second stage of inversion specifically involves: fixing the parameters obtained from the first stage of inversion and performing inversion using observation data of functional near-infrared spectral signals.

[0092] Furthermore, the functional near-infrared spectral signal observation model includes a tandem neurovascular coupling model and a blood oxygen dynamics model, wherein the neurovascular coupling model is used to convert neural activity into vasodilation signals, and its coupling efficiency parameter is one of the key parameters to be inverted.

[0093] The specific implementation of the two-stage inversion is as follows:

[0094] The first stage of inversion involves fixing the parameters of the functional near-infrared spectral signal observation model as its prior mean. Using preprocessed EEG and surface electromyography (EMG) signal observation data, a variational Laplace algorithm is employed to update some or all parameters of the neuronal dynamics model (especially connectivity parameters related to motor pathways) as well as the parameters of the EEG and EMG signal observation models. By maximizing the negative variational free energy (F) of the model, the mean (μ) and covariance (Σ) of the posterior distribution of the parameters are iteratively updated until convergence, yielding the posterior estimate for the first stage.

[0095] The update formula for the posterior mean of the parameters is as follows:

[0096] ;

[0097] in, Let be the posterior mean vector of the parameters in the k-th iteration, representing the optimal estimate of the model parameters (such as connection weights, coupling efficiency, etc.) in the current iteration; Let be the posterior mean vector of the parameters after the (k+1)th iteration, representing the new estimate obtained by adjusting the current mean; Let F be the gradient vector (first derivative) of the free energy F with respect to the mean parameter μ. The gradient represents the direction of change of the free energy in the parameter space, that is, the direction in which the free energy increases the most. In the inversion, maximizing the free energy is equivalent to making the gradient approach zero. Let F be the Hessian matrix (second derivative matrix) of the free energy F, which represents the curvature of the free energy in the parameter space, i.e. the rate of change of the gradient. It is the inverse of the negative Hessian matrix, used to scale the gradient step size to ensure that the update direction is correct and the step size is reasonable.

[0098] It should be noted that the update process of the posterior mean of the parameters adjusts the current mean in the direction of maximum free energy by combining gradient and curvature information. The iteration continues until the change in the mean converges (i.e., the gradient approaches zero).

[0099] The update formula for the posterior covariance of the parameters is as follows:

[0100] ;

[0101] in, Let be the posterior covariance matrix of the parameters in the (k+1)th iteration, representing the uncertainty of the updated parameters or the correlation between the parameters. Let F be the Hessian matrix (second derivative matrix) of the free energy F, which represents the curvature of the free energy in the parameter space, i.e. the rate of change of the gradient. The covariance matrix is ​​the inverse of the negative Hessian matrix. In Bayesian inference, when the posterior distribution is approximated as a Gaussian distribution, its covariance matrix is ​​usually negatively inversely correlated with the log-posterior Hessian matrix.

[0102] The second stage of inversion involves fixing all parameters obtained in the first stage (neuronal dynamics parameters, EEG signal, and surface electromyography signal observation model parameters). Using the preprocessed functional near-infrared spectral signal observation data, the variational Laplace algorithm is also employed to refine the functional near-infrared spectral signal observation model parameters (especially the neurovascular coupling efficiency parameter). Inversion estimation is performed to finally obtain the joint posterior distribution of all parameters.

[0103] It should be noted that the update method in the second stage is similar to that in the first stage, and will not be repeated here.

[0104] S40. Based on the model parameters obtained by inversion, generate the biomarkers of the subject.

[0105] In this embodiment, based on the posterior distribution (e.g., mean) of the parameters obtained from the final inversion, parameters with clear physiological significance are extracted, such as brain region connectivity strength parameters (i.e., exogenous connectivity parameters) and neurovascular coupling efficiency parameters (…). Neuromuscular conduction delay (D), etc., serve as interpretable biomarkers for quantifying coupling relationships between systems and are used in clinical motor function assessment.

[0106] Optionally, the multimodal physiological signal coupling analysis method further includes a model comparison and selection step, specifically:

[0107] Based on different physiological hypotheses (e.g., task modulation occurs only in the prefrontal-to-sensory-motor connection, or only in the output connection from the sensorimotor area to the spinal cord), multiple candidate multimodal dynamic causal models are constructed. The inversion process described in S30 above is performed on each candidate multimodal dynamic causal model, and its respective model evidence (e.g., negative variational free energy F) is calculated. The candidate multimodal dynamic causal model with the highest model evidence is selected as the optimal multimodal dynamic causal model, and S40 is performed based on its inversion results to generate more reliable biomarkers.

[0108] Optionally, the multimodal physiological signal coupling analysis method further includes a model validation step, specifically:

[0109] Using the optimal multimodal dynamic causal model and its posterior mean parameters, predicted time series of EEG, surface electromyography (EMG), and functional near-infrared spectral signals are generated through forward simulation. The predicted signals are then graphically overlaid with the actual observation data for comparison, providing an intuitive assessment of the model's goodness of fit.

[0110] The beneficial effects of this invention are as follows: By employing a staged Bayesian inversion strategy, it effectively solves the technical challenges of high computational complexity and unrobust parameter estimation caused by large differences in the time scales of multimodal data and numerous model parameters. Specifically, this invention first decomposes the joint inversion problem of high-dimensional parameter space into two consecutive low-dimensional inversion stages, thereby significantly reducing computational complexity and enabling the efficient and stable operation of complex models on ordinary computing devices. Furthermore, the staged strategy avoids mutual interference between parameters of different properties, making the parameter update process more stable and less prone to getting trapped in local optima, thus significantly improving the robustness and accuracy of parameter estimation. On this basis, this invention achieves deep fusion of multimodal information at the unified biophysical model level rather than simple superposition, making full use of the precise temporal information of high-frequency signals (such as EEG signals or surface electromyography signals) and the metabolic verification information of low-frequency signals (such as functional near-infrared spectroscopy signals), forming a complementary advantage. Ultimately, by using model parameters with clear physiological significance obtained from the inversion (such as connection strength parameters and coupling efficiency) to quantify the coupling relationship between systems, quantifiable and interpretable objective biomarkers are generated, providing a direct and reliable quantitative tool for clinical evaluation and effectively solving the problem of the lack of interpretable biomarkers in existing technologies.

[0111] Furthermore, such as Figure 3 As shown, based on the above-described multimodal physiological signal coupling analysis method, the present invention also provides a multimodal physiological signal coupling analysis system, the multimodal physiological signal coupling analysis system comprising:

[0112] The signal acquisition module 51 is used to simultaneously acquire physiological signals of the subject in multiple modalities when performing a specific experimental paradigm;

[0113] The model building module 52 is used to build a multimodal dynamic causal model, which includes a neuronal dynamics model and observation models corresponding to various physiological signals respectively.

[0114] The inversion estimation module 53 is used to invert and estimate the parameters of the multimodal dynamic causal model based on the physiological signal using a staged Bayesian inversion strategy, so as to obtain the model parameters.

[0115] Biomarker generation module 54 is used to generate biomarkers for the subject based on the model parameters obtained by inversion.

[0116] Furthermore, such as Figure 4 As shown, based on the above-mentioned multimodal physiological signal coupling analysis method and system, the present invention also provides a terminal, which includes a processor 10, a memory 20 and a display 30. Figure 4Only some of the terminal components are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0117] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory. In other embodiments, the memory 20 may be an external storage device of the terminal, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc. Further, the memory 20 may include both internal and external storage devices. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code installed on the terminal. The memory 20 can also be used to temporarily store data that has been output or will be output. In one embodiment, the memory 20 stores a multimodal physiological signal coupling analysis program 40, which can be executed by the processor 10 to implement the multimodal physiological signal coupling analysis method of this application.

[0118] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor, or other data processing chip, used to run program code stored in the memory 20 or process data, such as executing the multimodal physiological signal coupling analysis method.

[0119] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the terminal and to display a visual user interface. The components of the terminal communicate with each other via a system bus.

[0120] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a multimodal physiological signal coupling analysis program, which, when executed by a processor, implements the steps of the multimodal physiological signal coupling analysis method as described above.

[0121] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal that includes that element.

[0122] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0123] It should be understood that the application of the present invention is not limited to the examples above. Those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method of multi-modal physiological signal coupling analysis, characterized in that, The multi-modal physiological signal coupling analysis method comprises: synchronously collecting multiple modal physiological signals of a subject when performing a specific experimental paradigm; constructing a multi-modal dynamic causal model, the multi-modal dynamic causal model comprising a neuron dynamics model and observation models corresponding to the various physiological signals respectively; performing inversion estimation on parameters of the multi-modal dynamic causal model according to the physiological signals by using a staged Bayesian inversion strategy to obtain model parameters; generating biomarkers of the subject based on the inversion-obtained model parameters; wherein the multiple modal physiological signals comprise a first type of physiological signal and a second type of physiological signal; the performing inversion estimation on parameters of the multi-modal dynamic causal model according to the physiological signals by using a staged Bayesian inversion strategy to obtain model parameters comprises: fixing the observation model parameters corresponding to the first type of physiological signal, and performing inversion estimation on part or all of the parameters of the neuron dynamics model and the observation model parameters of the second type of physiological signal by using observation data of the second type of physiological signal; fixing the parameters obtained by the first-stage inversion estimation, and performing inversion estimation on the observation model parameters of the first type of physiological signal by using observation data of the first type of physiological signal to obtain a joint posterior distribution of all parameters; the first type of physiological signal is a low-frequency signal with a time resolution lower than a preset threshold, and the second type of physiological signal is a high-frequency signal with a time resolution higher than the preset threshold; wherein the low-frequency signal comprises a functional near-infrared spectroscopy signal, and the high-frequency signal comprises an electroencephalogram signal and a surface electromyogram signal; the staged Bayesian inversion strategy uses a variational Laplace algorithm which maximizes the negative variational free energy of the model by iteratively updating the mean and covariance of the parameter posterior distribution; the updating formula of the mean of the parameter posterior distribution is as follows: ； in, Let be the posterior mean vector of the parameters in the k-th iteration, representing the optimal estimate of the model parameters in the current iteration; Let be the posterior mean vector of the parameters after the (k+1)th iteration, representing the new estimate obtained by adjusting the current mean; The free energy F is relative to the mean of the parameters. the updating formula of the covariance of the parameter posterior distribution is as follows: gradient vector, Let F be the Hessian matrix of the free energy F, and let F be the curvature of the free energy in the parameter space. It is the inverse of the negative Hessian matrix, used to scale the gradient step size; the neuron dynamics model is a canonical microcircuit neuron population model used to simulate neural electrical activity within and between brain regions. ； wherein, is the parameter posterior covariance matrix for the k+1th iteration, representing the uncertainty of the updated parameters or the correlation between parameters.

2. The multi-modal physiological signal coupling analysis method of claim 1, wherein, The canonical microcircuit neuron population model comprises four neuron cell populations distributed in the granular layer, the upper granular layer and the lower granular layer, the four neuron cell populations comprising an excitatory multi-spiked stellate cell population, a superficial pyramidal cell population, a deep pyramidal cell population and an inhibitory interneuron cell population, the neuron cell populations being connected within a brain region through endogenous connection parameters, and the models of different brain regions being connected between brain regions through exogenous connection parameters.

3. The multi-modal physiological signal coupling analysis method of claim 2, wherein, the observation models corresponding to the various physiological signals respectively comprise:

4. The multi-modal physiological signal coupling analysis method of claim 1, wherein, an electroencephalogram signal observation model used to convert transmembrane potentials generated by neuron activity into scalp potential signals through a linear mapping matrix; a functional near-infrared spectroscopy signal observation model used to describe that vasodilation caused by neuron activity leads to changes in cerebral blood flow and drives changes in hemoglobin concentration; a surface electromyogram signal observation model used to convert motor cortex descending nerve driving signals into electromyogram signals recorded on the skin surface; ​ The to-be-estimated parameters of the surface electromyography signal observation model include a basis function weight parameter and a neuromuscular conduction delay parameter.

5. A multi-modal physiological signal coupling analysis system, characterized by, The multi-modal physiological signal coupling analysis system is used to implement the steps of the multi-modal physiological signal coupling analysis method according to any one of claims 1-4, and the multi-modal physiological signal coupling analysis system comprises: a signal acquisition module configured to synchronously acquire a plurality of modal physiological signals of a subject when the subject performs a specific experimental paradigm; a model construction module configured to construct a multi-modal dynamic causal model, the multi-modal dynamic causal model comprising a neuron dynamics model and an observation model corresponding to each of the physiological signals; an inversion estimation module configured to perform inversion estimation on parameters of the multi-modal dynamic causal model according to the physiological signals by using a staged Bayesian inversion strategy, to obtain model parameters; a marker generation module configured to generate a biomarker of the subject based on the inversion-obtained model parameters.

6. A terminal, characterized by comprising: The terminal comprises a memory, a processor, and a multi-modal physiological signal coupling analysis program stored on the memory and executable on the processor, and the multi-modal physiological signal coupling analysis program, when executed by the processor, implements the steps of the multi-modal physiological signal coupling analysis method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a multi-modal physiological signal coupling analysis program, and the multi-modal physiological signal coupling analysis program, when executed by a processor, implements the steps of the multi-modal physiological signal coupling analysis method according to any one of claims 1-4.

Images